Amazon reported second-quarter earnings on Thursday that sent its stock soaring nearly 10% in after-hours trading, driven by a 37% year-over-year jump in cloud revenue to $42 billion. The results defied conventional wisdom that investors want companies to rein in data center spending, as Amazon revealed it is actually accelerating its massive infrastructure buildout. Net sales rose 20% in the quarter, but the real story lies in the company's escalating capital expenditures and the broader implications for the entire artificial intelligence economy.
Cloud Revenue Soars, But Capex Balloons to $173 Billion
Amazon Web Services (AWS) revenue was the undisputed bright spot in Q2, growing 37% year over year to $42 billion. Yet the company is spending at a pace that far outstrips even that growth. For the fiscal year ended June 30, Amazon spent $173 billion on property and equipment, a category that includes GPUs, natural gas turbines, and plots of land. That is up sharply from $107.65 billion in the previous fiscal year. The company also raised its 2026 capital expenditure forecast from $200 billion to $220 billion, signaling no intention to slow down despite investor sentiment that has punished other tech giants for similar spending.
The spending is so aggressive that Amazon began dipping into its cash reserves to cover costs. The company ended the quarter with $7.6 billion less cash than it had 12 months ago, marking the first period of negative free cash flow this year. The raw arithmetic suggests that even AWS revenue is not enough to balance the capex, but Amazon is betting that demand will continue to grow alongside supply.
Jassy Defends the Buildout: Margins, Models, and Long-Term Bets
On the Q2 earnings call, Amazon CEO Andy Jassy addressed the spending directly. "We see the AI business following very much the same margin trajectory we saw in the core business before," he said, suggesting that the massive upfront investment will eventually pay off as it did for Amazon's e-commerce and cloud operations. He also pushed back on the idea that Amazon needs to own a frontier AI model to succeed in the space. "AWS and Amazon Bedrock can have a wildly successful business without its own frontier model, and the reason is that there's not going to be a single model to rule them all," Jassy said.
Amazon is making long-term bets on its own chip projects, including the Trainium TPU and Arm-based Graviton processor. Those projects do not show up in the capex numbers, but they can meaningfully improve margins for the cloud business over time. There is a years-long time lag between breaking ground on a data center and selling its capacity, meaning today's spending is a bet on demand that may not materialize for years.
Investors Treat Cloud Hosts as the Safe Bet in AI
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Amazon's stock surge mirrors similar pops at Microsoft and Google after they reported strong cloud revenue. All three cloud hosts are being treated by investors as the most reliable part of the AI stack. Meta, by contrast, saw its stock fall 8% after earnings this week. The social media company has significant capital expenditure on AI but no clear revenue source to show for it, leaving investors skeptical due to a cash flow crunch and continued spending.
The pattern is clear: investors love AI as long as you are a cloud host. They are far more skeptical about the underlying economics for AI labs and AI startups. Amazon's hosting revenue is someone else's AI bill, for example, Anthropic, the AI lab that runs much of its workload on AWS. If AI lab spending is not sustainable, then cloud host revenue will not be stable either.
The $3 Trillion Question: Will AI Demand Hold Up?
The article references a question posed by analyst David Cahn: the "$3 trillion question" of whether AI demand will justify the massive buildout underway. If demand for AI does not hold up, it will be a bad time for everyone in the stack. Cloud hosts may be somewhat insulated from the demand problem, but not completely. If AI lab spending collapses, the hosting revenue that investors are currently rewarding will evaporate.
The numbers are staggering. Amazon spent $173 billion on property and equipment in the fiscal year ended June 30, up from $107.65 billion the year before. It raised its 2026 capex forecast from $200 billion to $220 billion. It ended the quarter with $7.6 billion less cash than 12 months ago. AWS revenue hit $42 billion in Q2, up 37% year over year. Net sales rose 20%. The stock popped nearly 10% after-hours. Meta fell 8% this week.
But the $3 trillion question remains unanswered. The buildout is a bet that demand will continue to grow at a pace that justifies spending that already exceeds $173 billion a year. If that bet fails, the consequences will ripple through the entire AI economy, from cloud hosts to AI labs to the startups that depend on them. For now, investors are betting that Amazon, Microsoft, and Google are the safest places to be. But the math is not yet settled.